Capstok — learn by doing

Q1.Order these voice-turn states as a single successful turn should pass through them.

rank

Re-order from top (rank 1) to bottom.

1.Listening

2.Speaking

3.Thinking

4.Idle

Q2.Match each build-vs-buy consideration to the stack it favors.

match

Match each item on the left to one on the right.

Need fine-grained control over tool-calling text between stages

↔

Latency is the single top priority and tool control isn't critical

↔

Need to swap providers later for cost or quality reasons

↔

Q3.Select every true statement about streaming versus batch voice pipelines.

multi-select

Q4.Why does sending audio at the wrong sample rate produce sped-up or slowed-down speech rather than an outright error?

single

Q5.Roughly what average gap between turns does human conversation research use as the 'feels natural' benchmark?

single

Q6.What are the three stages of the classic assembled voice-agent loop, in order?

single

Q7.The metric that captures your worst 5% of turns — the ones users actually remember as laggy — is called the ______.

fill in

Quiz me — The Realtime Voice Stack